Simple regression models
Fitting and assessing simple models
Jonathan Binder
Royal Holloway, University of London
16 Feb 2026
Statistical models
- mathematical representation of phenomena
- models consist of dependent variable (outcome) and one or more independent variable (predictor)
Linear model assumptions
- linear relationship between predictor and outcome - an increase in x is associated with a proportional increase in y no matter the value of x
Linear model assumptions
- independence of errors (no repeated measures/time series, no spatial autocorrelation, etc.)
- homoscedasticity (constant variance) of the errors - the variance of the errors is the same across all values of x
- errors are normally distributed